Re: Superko detection (was Re: [computer-go] M.Boon-Library)

[Gunnar Farnebäck <gunnar@xxxxxxxxxxxxxx>]

Frank wrote:
> I thought that when doing MadLab- / GoTools-like L&D search, that SuperKo 
> was important tot check for as pruning contraint? When generating the 
> movetree, would you otherwise not generate a lot of superko situations?

If you are doing more or less exhaustive search with the ambition to
actually prove something some kind of strict detection of repetition
is almost certainly necessary. If you instead do relatively narrow
searches with depth limits, as is generally the case in GNU Go, things
are easier. The worst that can happen is that you loop around a few
turns but most of that can be avoided by heuristics in the move
generation. For example GNU Go is very restrictive in allowing
repeated ko captures in multiple kos. More so than what superko
requires.

/Gunnar
_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/